CN109872318A - A kind of geology for deep learning is appeared crack data set production method - Google Patents
A kind of geology for deep learning is appeared crack data set production method Download PDFInfo
- Publication number
- CN109872318A CN109872318A CN201910131407.0A CN201910131407A CN109872318A CN 109872318 A CN109872318 A CN 109872318A CN 201910131407 A CN201910131407 A CN 201910131407A CN 109872318 A CN109872318 A CN 109872318A
- Authority
- CN
- China
- Prior art keywords
- crack
- geology
- appeared
- sample
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
It appears crack data set production method the invention discloses a kind of geology for deep learning, basic step are as follows: the original image appeared using the geology that unmanned plane obtains objective area；Determine the size in sample constituency；In the selection for the enterprising row crop crack sample of original image that geology is appeared；The selection of non-crack of appearing (complex background) sample is carried out on the original image that geology is appeared；To appearing crack sample and non-crack sample of appearing is marked；Quality evaluation is carried out to the sample data set that has marked, the geology that high quality finally can be obtained is appeared crack data set.Method provided by the invention have many advantages, such as it is scientific and reasonable, be easily achieved, quality it is high, by the way that the appear classification of middle complex background rock mass of geology is summarized and chosen, the geology for producing high quality is appeared crack data set.
Description
Technical field
The present invention relates to image generate technical field more particularly to geology appear crack image pattern generation application neck
Domain, specifically a kind of geology for deep learning are appeared crack data set production method.
Background technique
The observation that geology is appeared is the essential research contents of geologist all the time, this work is for a long time
It needs geological personnel to reach objective area, research object is taken pictures or Freehandhand-drawing record is next for statistical analysis, this mode works
It is low and have some potential safety problems that low efficiency, some areas can not obtain data, data user rate.With the hair of deep learning
Exhibition, the recognition capability of model is more and more stronger, has successfully realized the identification to crack in road and bridge.Therefore, it appears and splits to geology
The automatic identification of seam has become possibility, and has great practical value.
The appear first step of crack automatic recognition of geology is exactly the sample data set with high quality, and compares road and bridge
In crack sample, the appear sample in crack of geology has a more complicated background, and identification difficulty is bigger.This method is to complicated back
Scape rock mass is classified summary and selection, realizes high quality geology and appears the production of crack sample data set.
Summary of the invention
(1) technical problems to be solved
It appears crack data set production method, appears to geology multiple the present invention provides a kind of geology for deep learning
Miscellaneous geologic setting is classified summary and selection, and having made can be used for the geology of deep learning model and appear crack sample
Data set.
(2) technical solution
The present invention comprises the steps of:
(1) it is appeared using geology of the unmanned plane to objective area and carries out data acquisition, obtain the original image that geology is appeared；
(2) sample constituency size is determined；
(3) size in the sample constituency according to determined by step (2) is appeared original in step (1) geology obtained
On image, crack sample selection of appearing is carried out；
(4) size in the sample constituency according to determined by step (2) is appeared original in step (1) geology obtained
On image, carries out non-crack of appearing (complex background) sample and choose；
(5) to selected in step (3) and step (4) appear crack sample and non-crack of appearing (complex background) sample into
Line flag obtains tag image；
(6) deep learning model is utilized, quality evaluation is carried out to the sample data set marked.
Further, resolution ratio is better than 1cm when data acquire in the step (1), and picture format is 2-D color image
(.jpg)。
Further, sample constituency size is 256 × 256 (units: pixel) in the step (2).
Further, non-crack of appearing (complex background) includes: that massif and massif, massif and sky are handed in the step (4)
Boundary, Artificial facilities, surface water current mark (include: containing sedimentation, salt alkali precipitation, scour mark) massif gravel buildup body, shade side
Boundary, karst zone, surface layer weathering seam.
Further, in the step (5), non-crack of appearing (complex background) is labeled as 0, and crack of appearing is marked labeled as 1
The format of image is 2-D gray level image (.png).
Further, deep learning model used is SegNet, GoogleNet, UNet, quality evaluation in the step (6)
Index is nicety of grading, and is all higher than 85%.
(3) beneficial effect
Advantages of the present invention is embodied in:
Since geology is appeared complicated background rock mass, very big choose is brought for the appear automatic identification in middle crack of geology
War.The present invention to geology appear middle different geologic body carried out summarize classification and choose, specify geology appear in non-crack
The type of complex background rock mass, the geology for having produced high quality are appeared crack data set, are appeared the automatic of crack for geology
Identification has great importance.
Detailed description of the invention
Fig. 1 is the step flow chart that the present invention is implemented,
Fig. 2 is the schematic diagram for choosing crack sample of appearing on the original image,
Fig. 3 is that massif and massif, the massif and sky in the non-crack sample of appearing chosen have a common boundary,
Fig. 4 is the Artificial facilities in the non-crack sample of appearing chosen,
Fig. 5 is the surface water current mark in the non-crack sample of appearing chosen,
Fig. 6 is the massif gravel buildup body in the non-crack sample of appearing chosen,
Fig. 7 is the shadow edge in the non-crack sample of appearing chosen,
Fig. 8 is the karst zone in the non-crack sample of appearing chosen,
Fig. 9 is the surface layer weathering seam in the non-crack sample of appearing chosen,
Figure 10 is sample labeling schematic diagram.
Specific embodiment
To keep the purpose of the present invention, content and advantage clearer, with reference to the accompanying drawings and examples, to of the invention
Specific embodiment is described in further detail:
Referring to Fig.1, specific implementation step of the invention are as follows:
(1) it is appeared using geology of the unmanned plane to objective area and carries out data acquisition, obtain the original image that geology is appeared；
Wherein, resolution ratio is better than 1cm when data acquire, and picture format is 2-D color image (.jpg).
(2) sample constituency size is determined；
Wherein, sample constituency size is 256 × 256 (units: pixel).
(3) referring to Fig. 2, according to the size in sample constituency determined by step (2), in step (1) geology dew obtained
On the original image of head, crack sample selection of appearing is carried out.
(4) size in the sample constituency according to determined by step (2) is appeared original in step (1) geology obtained
On image, carries out non-crack of appearing (complex background) sample and choose.
Wherein, non-crack of appearing (complex background) includes: that massif and massif, massif and sky have a common boundary (such as Fig. 3), is manually set
It applies (such as Fig. 4), surface water current mark (includes: containing sedimentation, salt alkali precipitation, scour mark) (such as Fig. 5), massif gravel buildup body
(such as Fig. 9) is stitched in (such as Fig. 6), shadow edge (such as Fig. 7), karst zone (such as Fig. 8), surface layer weathering.
(5) referring to Fig.1 0, to appear crack sample and the non-crack of appearing selected in step (3) and step (4), (complexity is carried on the back
Scape) sample is marked, obtain tag image；
Wherein, non-crack of appearing (complex background) is labeled as 0, and labeled as 1, the format of tag image is 2-D in crack of appearing
Gray level image (.png).
(6) deep learning model is utilized, quality evaluation is carried out to the sample data set marked；
Wherein, deep learning model used is SegNet, GoogleNet, UNet, and quality evaluation index is nicety of grading,
And all it is higher than 85%.
Claims (6)
- The crack data set production method 1. a kind of geology for deep learning is appeared, which comprises the following steps:(1) it is appeared using geology of the unmanned plane to objective area and carries out data acquisition, obtain the original image that geology is appeared；(2) sample constituency size is determined；(3) size in the sample constituency according to determined by step (2), in the original image that step (1) geology obtained is appeared On, carry out crack sample selection of appearing；(4) size in the sample constituency according to determined by step (2), in the original image that step (1) geology obtained is appeared On, it carries out non-crack of appearing (complex background) sample and chooses；(5) to selected in step (3) and step (4) appear crack sample and non-crack of appearing (complex background) sample is marked Note, obtains tag image；(6) deep learning model is utilized, quality evaluation is carried out to the sample data set marked.
- The crack data set production method 2. a kind of geology for deep learning according to claim 1 is appeared, feature Be: for resolution ratio better than 1cm, picture format is 2-D color image (.jpg) when data acquire in the step (1).
- The crack data set production method 3. a kind of geology for deep learning according to claim 1 is appeared, feature Be: sample constituency size is 256 × 256 (units: pixel) in the step (2).
- The crack data set production method 4. a kind of geology for deep learning according to claim 1 is appeared, feature Be: non-crack of appearing (complex background) includes: that massif and massif, massif and sky have a common boundary in the step (4), is manually set It applies, surface water current mark (includes: containing sedimentation, salt alkali precipitation, scour mark) massif gravel buildup body, shadow edge, weathering leaching Strainer, surface layer weathering seam.
- The crack data set production method 5. a kind of geology for deep learning according to claim 1 is appeared, feature Be: in the step (5), non-crack of appearing (complex background) is labeled as 0, and crack of appearing is labeled as 1, the format of tag image For 2-D gray level image (.png).
- The crack data set production method 6. a kind of geology for deep learning according to claim 1 is appeared, feature Be: deep learning model used is SegNet, GoogleNet, UNet in the step (6), and quality evaluation index is classification Precision, and all it is higher than 85%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910131407.0A CN109872318A (en) | 2019-02-22 | 2019-02-22 | A kind of geology for deep learning is appeared crack data set production method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910131407.0A CN109872318A (en) | 2019-02-22 | 2019-02-22 | A kind of geology for deep learning is appeared crack data set production method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109872318A true CN109872318A (en) | 2019-06-11 |
Family
ID=66919031
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910131407.0A Pending CN109872318A (en) | 2019-02-22 | 2019-02-22 | A kind of geology for deep learning is appeared crack data set production method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109872318A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569730A (en) * | 2019-08-06 | 2019-12-13 | 福建农林大学 | Road surface crack automatic identification method based on U-net neural network model |
| CN112307803A (en) * | 2019-07-25 | 2021-02-02 | 中国石油天然气股份有限公司 | Digital geological outcrop crack extraction method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101772916B1 (en) * | 2016-12-30 | 2017-08-31 | 한양대학교 에리카산학협력단 | Device for measuring crack width of concretestructure |
| CN107480611A (en) * | 2017-07-31 | 2017-12-15 | 浙江大学 | A kind of crack identification method based on deep learning convolutional neural networks |
| CN108364280A (en) * | 2018-01-03 | 2018-08-03 | 东南大学 | Structural cracks automation describes and width accurately measures method and apparatus |
| CN108520516A (en) * | 2018-04-09 | 2018-09-11 | 陕西师范大学 | A kind of bridge pavement Crack Detection and dividing method based on semantic segmentation |
| CN109102029A (en) * | 2018-08-23 | 2018-12-28 | 重庆科技学院 | Information, which maximizes, generates confrontation network model synthesis face sample quality appraisal procedure |
| CN109146849A (en) * | 2018-07-26 | 2019-01-04 | 昆明理工大学 | A kind of road surface crack detection method based on convolutional neural networks and image recognition |
-
2019
- 2019-02-22 CN CN201910131407.0A patent/CN109872318A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101772916B1 (en) * | 2016-12-30 | 2017-08-31 | 한양대학교 에리카산학협력단 | Device for measuring crack width of concretestructure |
| CN107480611A (en) * | 2017-07-31 | 2017-12-15 | 浙江大学 | A kind of crack identification method based on deep learning convolutional neural networks |
| CN108364280A (en) * | 2018-01-03 | 2018-08-03 | 东南大学 | Structural cracks automation describes and width accurately measures method and apparatus |
| CN108520516A (en) * | 2018-04-09 | 2018-09-11 | 陕西师范大学 | A kind of bridge pavement Crack Detection and dividing method based on semantic segmentation |
| CN109146849A (en) * | 2018-07-26 | 2019-01-04 | 昆明理工大学 | A kind of road surface crack detection method based on convolutional neural networks and image recognition |
| CN109102029A (en) * | 2018-08-23 | 2018-12-28 | 重庆科技学院 | Information, which maximizes, generates confrontation network model synthesis face sample quality appraisal procedure |
Non-Patent Citations (1)
| Title |
|---|
| 程海燕等: "无人机航拍电力部件图像准确分类", 《计算机仿真》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112307803A (en) * | 2019-07-25 | 2021-02-02 | 中国石油天然气股份有限公司 | Digital geological outcrop crack extraction method and device |
| CN110569730A (en) * | 2019-08-06 | 2019-12-13 | 福建农林大学 | Road surface crack automatic identification method based on U-net neural network model |
| CN110569730B (en) * | 2019-08-06 | 2022-11-15 | 福建农林大学 | Road surface crack automatic identification method based on U-net neural network model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110276402B (en) | Salt body identification method based on deep learning semantic boundary enhancement | |
| CN109523566A (en) | A kind of automatic division method of Sandstone Slice micro-image | |
| CN109740608A (en) | A kind of image partition method based on deep learning | |
| CN113703045B (en) | Seismic facies identification method based on lightweight network | |
| CN109872318A (en) | A kind of geology for deep learning is appeared crack data set production method | |
| CN105093289A (en) | Seismic horizon identification method based on image processing | |
| CN109800728A (en) | A kind of minerals based on deep learning quickly know method for distinguishing | |
| Permana et al. | Analysis of microfacies and depositional environment of limestone in Yosonegoro area, Gorontalo Province, Indonesia | |
| CN112329706A (en) | Mining land identification method based on remote sensing technology | |
| CN107886125A (en) | MODIS satellite remote sensing images mask methods based on local spectral factorization marking | |
| Chavasseau et al. | Chapter 19. Advances in the biochronology and biostratigraphy of the continental Neogene of Myanmar | |
| STONE | WAULSORTIAN-TYPE BIOHERMS (REEFS) OF MISSISSIPPIAN AGE, CENTRAL BRIDGER RANGE, MONTANA1 | |
| CN112784806A (en) | Lithium-containing pegmatite vein extraction method based on full convolution neural network | |
| CN105719296B (en) | The high speed image two-value connected component labeling method indicated based on address-event | |
| Schetselaar et al. | Integrated geological interpretation of remotely sensed data to support geological mapping in Mozambique | |
| CN113780117B (en) | Method for rapidly identifying and extracting relevant parameters of estuary plume outline | |
| CN114743103A (en) | Island reef remote sensing image geological classification method based on Deeplabv3+ network model | |
| CN109285167B (en) | Method for extracting colored drawing pottery pattern based on deep learning | |
| Wolf | Conceptual models, 1. Examples in sedimentary petrology, environmental and stratigraphic reconstructions, and soil, reef, chemical and placer sedimentary ore deposits | |
| Ali | High-resolution stratigraphic architecture and sedimentological heterogeneity within the Miocene dam formation, Eastern Province, eastern Saudi Arabia | |
| Song et al. | Review of Nodule Mineral Image Segmentation Algorithms for Deep-Sea Mineral Resource Assessment. | |
| Bajwa et al. | A study for prediction of minerals in rock images using back propagation neural networks | |
| CN115471713B (en) | Shale strawberry-shaped pyrite particle size measurement method based on convolutional neural network | |
| Wolf | Conceptual Models in Geology. | |
| CN104182730A (en) | Automatic identification method of granite microsection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190611 |
|
| RJ01 | Rejection of invention patent application after publication |